Bourbaki

Appendix 3 hilbert’s nullstellensatz

Theorem 1 00QM

Let A be an integral domain, K be its field of fractions, and L be a commutative K-algebra. Suppose that the A-algebra L is generated by finitely many elements and that L is a field.

a) The degree of L over K is finite.

b) There exists a nonzero element aa of A such that K is equal to A[a−1]A[a^{-1}].

Let S be a generating subset of the A-algebra L. We use induction on the cardinal of S.

First, suppose that the elements of S are all algebraic over K. The degree of L over K is then finite (V, §3, No. 2, p. 18, Theorem 2). Let (ei)i∈I(e_i)_{i\in I} be a basis of L over K. There exists a nonzero element aa of A such that the coordinates with respect to this basis of the elements of S, of the element 1, and of the elements eieje_ie_j for i,ji,j in I belong to A[a−1]A[a^{-1}]. The set of linear combinations ∑i∈Iaiei\sum_{i\in I}a_ie_i with ai∈A[a−1]a_i\in A[a^{-1}] for every ii is then a subring of L. It contains A and S by construction and is therefore equal to L. In particular, it contains Ke1Ke_1, so K is equal to A[a−1]A[a^{-1}].

Now suppose that an element ss of S is transcendental over K. Denote by E the field of fractions of the ring A[s]A[s]. The A[sA[s]-algebra L is generated by S −{s}-\{s\}. By the induction hypothesis, there exists a nonzero polynomial P∈A[X]P\in A[X] such that E is equal to A[s][P(s)−1]A[s][P(s)^{-1}]. Let K‾\overline{K} be an algebraic closure of K (V, §4, No. 3, p. 23, Theorem 2). Since the field K‾\overline{K} is infinite (V, §4, No. 1, p. 20, Proposition 3), there exists an element xx of K‾\overline{K} such that P(x)≠0P(x)\not= 0. Let φ:E→K\varphi : E\rightarrow K be the unique homomorphism from the K-algebra E=A[s][P(s)−1]E = A[s][P(s)^{-1}] to the K-algebra K‾\overline{K} that sends ss to xx. This is absurd because E is a transcendental extension of K and K‾\overline{K} is an algebraic extension of K. This completes the proof of the theorem.

Corollary 1 00RV

Let K be a commutative field, A be a commutative K-algebra generated by finitely many elements, and m\mathfrak{m} be a maximal ideal of A.

a) The degree of A/mA/\mathfrak{m} over K is finite.

b) Let Ω be an algebraically closed extension of K. There exists a K-algebra homomorphism from A to Ω with kernel m\mathfrak{m}.

The K-algebra A/mA/\mathfrak{m} is generated by finitely many elements, and A/mA/\mathfrak{m} is a field. By Theorem 1, the degree of A/mA/\mathfrak{m} is finite; assertion a) follows. Every extension of K of finite degree is isomorphic to a subextension of Ω (V, §4, No. 1, p. 20, Theorem 1); this gives b).

Corollary 2 00QN

Let K be a commutative field, nn be a natural number, (Pi)i∈I(P_i)_{i\in I} be a family of elements of K[X1,...,Xn]K[X_1, . . . ,X_n], and Ω be an algebraically closed extension of K. The following conditions are equivalent:

(i) The polynomials PiP_i do not have a common zero in Ωn\Omega^n.

(ii) There exists a family (Qi)i∈I(Q_i)_{i\in I}, with finite support, of elements of K[X1,...,Xn]K[X_1, . . . ,X_n] such that ∑i∈IPiQi=1\sum_{i\in I}P_iQ_i= 1.

Denote by A the ring K[X1,...,Xn]K[X_1, . . . ,X_n] and by a\mathfrak{a} the ideal generated by the polynomial PiP_i. Condition (i) means that there are no K-algebra homomorphisms from A/aA/\mathfrak{a} to Ω. If it is satisfied, then by Corollary 1, the ring A/aA/\mathfrak{a} has no maximal ideals and is therefore zero, and 1 belongs to a\mathfrak{a}. This proves that (i) implies (ii). The implication (ii) ⇒\Rightarrow (i) is clear.

Transcribed from the printed edition, 2023, Springer Nature. Not translated.